Skip to Content
AI Agents (MCP)Limits and errors

Limits and errors

Rate limits

LimitValue
Per token60 units per minute
Per organization (all its tokens)300 units per minute

A tool call costs 1 unit; heavy tools (large installation lists, long analytics ranges, cross-area search) cost 3. Other requests, such as listing tools, cost 1. Over the limit, the server answers HTTP 429 with a Retry-After header (in seconds).

Time limits

  • A single query inside a tool stops after 5 seconds. Narrow it: a shorter date range, a filter, a smaller page.
  • One HTTP request may spend 15 seconds of tool time in total. When an agent batches several tool calls into one request, calls after that point return an “over budget” error; send them in a new request.

Result size

Lists return up to 100 items per page (default 25) with a next_cursor. A result over 50 KB is replaced by "truncated": true and a hint to narrow the query.

Errors

You seeMeaning
HTTP 401The token is missing, mistyped, expired or revoked, or its creator no longer has access.
HTTP 403The organization is suspended, the plan doesn’t include AI agents, or the request came from a web page on another origin.
HTTP 429Rate limited. Wait for Retry-After seconds. Repeated failed authentication from one address is also limited.
HTTP 503Appneck is in scheduled maintenance. Retry after Retry-After seconds.
Tool error “lacks the … scope”The token doesn’t have that scope, or your role no longer allows it.
Tool error “not found”No such item in this organization. Items in other organizations are always “not found”.
Tool error “took too long”The query hit the 5-second limit. Narrow it.